Bias reduction for stereo triangulation

Stereo triangulation lays at the basis of 3D scene recovery and it is used in a wide variety of areas ranging from urban modelling to robot localisation and mapping. However, triangulation produces non-Gaussian 3D estimates from Gaussian image measurements owing to its nonlinear nature. While previous work demonstrates the presence of statistical bias and how to correct the depth estimate, in this presented report, proposed and proven in a Monte Carlo test, is an enhancement for correcting the full 3D position given the image projection noise variance.